Offline voice control for lights: no account, no network, no cloud

Ask a smart speaker to turn the light on while the broadband is down and nothing happens. Some of our lights listen for spoken commands with no connection at all — the recognition runs on a chip a few centimetres from the LED. It is a different product from cloud voice, and the differences run in both directions.

Where the recognition happens decides everything

Cloud voice is what most people mean by voice control. You speak to a smart speaker or a phone, the audio goes to a server, the server works out what you meant, and a command comes back down to the light. Offline voice puts a small recognition engine inside the fitting instead. Everything else follows from that one difference.

Offline voice is not a cut-down version of the cloud approach. It is a different design with a different set of things it is good at, and it fails in different places.

What offline is better at

1. It answers immediately. Recognition happens where the light is, so there is no round trip to a server. The gap between finishing the phrase and the light changing is short enough to read as instant, which cloud voice usually does not.
2. It works when the network does not. The router reboots, the broadband drops, the account logs out on somebody's phone — none of it reaches the light. The moment the power comes back after a cut is when you most want a light to answer a voice, and it is exactly when a cloud assistant is least likely to.
3. Nothing leaves the room. There is no microphone stream going anywhere, because there is nowhere for it to go. For a bedroom, a child's room, or anyone who does not want an always-listening cloud device in the house, this is the entire argument.

What it gives up

1. The command set is fixed. You cannot invent a phrase. The list was decided at manufacture, and using the product means learning it rather than the other way round. Most sets run to a dozen or two commands — on, off, brighter, dimmer, colours, a few scenes.
2. It knows only itself. "Turn everything off" needs something that has a model of what "everything" is. An offline chip in one fitting has no idea the other fittings exist. Whole-house logic is what an assistant or a hub is for, and it is the thing offline voice structurally cannot do.
3. The wake word, where there is one, is not yours to change. Brands can choose it when a product is specified — a manufacturing option, and a common one on export models — but the person who bought the light cannot.

False triggers matter more than vocabulary

When comparing offline voice fittings, the specification people read is the command count, and the specification that decides whether they like the product is the false trigger rate. A fitting that wakes up during the television news is worse to live with than one that understands ten phrases instead of twenty.

Two figures describe this. The false acceptance rate is how often it responds when nobody addressed it. The false rejection rate is how often it ignores a command that was properly given. They trade against each other: tuning a fitting to be harder to trigger by accident also makes it harder to trigger on purpose. The useful question is what both numbers look like with background noise present, because the quiet-room figures are always good.

Where the microphone sits matters as much as the chip

A microphone behind a diffuser hears a muffled room. One facing the ceiling hears the ceiling. On a fitting mounted high in a room with hard floors and bare walls, the reflections arrive late enough to smear the phrase. None of this shows on a datasheet, which is why a fitting that recognises well on a desk can disappoint once it is installed. Test it the way it will live:

1. Hold or prop the fitting at its final height, in the room it will serve, before any screws go in.
2. Turn the television on at normal volume. Every voice product performs well in a quiet showroom; the living room is the test.
3. Give each command from where you would normally stand — the sofa, the doorway, the bed — not from arm's length.
4. Leave it listening for an evening and count how often it acts when nobody spoke to it. That count is the false acceptance rate for your room.

Languages are models, not settings

Offline recognition is per-language in a way cloud voice is not. Each language is a separate model compiled into the chip, so a product that understands English and Mandarin is carrying two, and adding a third is a manufacturing decision rather than a menu item. That is the constraint behind an oddity buyers notice: the same fitting sold in two markets may have different command sets, because the sets were built separately.

Model size is the other axis. A larger model copes better with accents and background noise and costs more silicon; a smaller one is cheaper and less forgiving. For a product sold into one market with one accent, small is often the right choice. For a product that will meet accents nobody tested against, it usually is not — and a household with two accents in it should test with both before the fitting goes up.

What the cloud assistants and Matter do when the internet is down

Since March 2025 Amazon has processed every Echo voice request in the cloud; the on-device option that a few Echo models offered was withdrawn. Google's Nest speakers have always sent speech to the cloud, with only alarms and timers working without a connection. A spoken command to a Govee or Hue light through either assistant therefore travels from your room to the assistant's servers, on to the light maker's servers, and back down to the light.

Matter changes the last hop, not the first. A Matter light can be controlled over the local network with the internet down, from a phone or a hub, and that is part of the specification. The speech still has to be recognised by the assistant, which needs the cloud. Standalone offline-voice lights do exist on the market, typically with a fixed list of around ten commands and no app at all; that fixed list is the same trade described above.

Sources: The Register, Amazon ends on-device Alexa processing; Connectivity Standards Alliance, Matter FAQ.

Choosing between them

Whichever you choose, run the four-step test above in the room before mounting. Which of our fittings carry offline voice, and which app pairs with the ones that do, is set out in which app controls your light. For ceiling fan lights, why fan lights run on mesh covers what a voice command reaches once the fitting is in a room group.

Questions about a specific fitting of ours? Contact us.